An essential, kinetoplastid-specific GDP-Fuc: β-D-Gal α-1,2-fucosyltransferase is located in the mitochondrion of Trypanosoma brucei

Fucose is a common component of eukaryotic cell-surface glycoconjugates, generally added by Golgi-resident fucosyltransferases. Whereas fucosylated glycoconjugates are rare in kinetoplastids, the biosynthesis of the nucleotide sugar GDP-Fuc has been shown to be essential in Trypanosoma brucei. Here we show that the single identifiable T. brucei fucosyltransferase (TbFUT1) is a GDP-Fuc: β-D-galactose α-1,2-fucosyltransferase with an apparent preference for a Galβ1,3GlcNAcβ1-O-R acceptor motif. Conditional null mutants of TbFUT1 demonstrated that it is essential for both the mammalian-infective bloodstream form and the insect vector-dwelling procyclic form. Unexpectedly, TbFUT1 was localized in the mitochondrion of T. brucei and found to be required for mitochondrial function in bloodstream form trypanosomes. Finally, the TbFUT1 gene was able to complement a Leishmania major mutant lacking the homologous fucosyltransferase gene (Guo et al., 2021). Together these results suggest that kinetoplastids possess an unusual, conserved and essential mitochondrial fucosyltransferase activity that may have therapeutic potential across trypanosomatids.


Introduction
The protozoan parasites of the Trypanosoma brucei group are the causative agents of human and animal African trypanosomiasis. Bloodstream form T. brucei is ingested by the tsetse fly vector and differentiates into procyclic form parasites to colonize the tsetse midgut. To then infect a new mammalian host, T. brucei undergoes a series of differentiations that allows it to colonize the fly salivary gland and to be transferred to a new host during a subsequent blood meal (Matthews, 2005).

Identification, cloning, and sequence analysis of TbFUT1
The CAZy database lists eight distinct FUT families (see Introduction) (Lombard et al., 2014). One or more sequences from each family were selected for BLASTp searches of the predicted proteins from the T. brucei, T. cruzi, and L. major genomes (Supplementary file 1). Strikingly, only one putative fucosyltransferase gene (TbFUT1) was identified in the T. brucei genome (GeneDB ID: Tb927.9.3600) belonging to the GT11 family, which is comprised almost exclusively of α-1,2-FUTs (Coutinho et al., 2003;Zhang et al., 2010). Homologues of TbFUT1 were also found in the T. cruzi and L. major genomes and, unlike T. brucei, T. cruzi, and L. major, also encode for GT10 FUT genes (Supplementary file 1). Finally, Leishmania spp. express a family of α-1,2-arabinopyranosyltransferases (SCA1/2/L, CAZy family GT79) that decorate phosphoglycan side chains and have been suggested to act as FUTs in presence of excess fucose (Guo et al., 2021).
Additional BLASTp searches showed that there is generally a single TbFUT1 gene homologue in each kinetoplastid species, and a phylogram of FUT sequences indicates that TbFUT1 homologues form a distinct clade closest to bacterial α-1,2-FUT ( Figure 1B).

Recombinant expression of TbFUT1
The TbFUT1 ORF was amplified from T. brucei 427 genomic DNA and cloned in the pGEX6P1 expression vector. The resulting construct (pGEX6P1-GST-PP-TbFUT1) encoded for the TbFUT1 ORF with a glutathione-S-transferase (GST) tag at its N-terminus and a PreScission Protease (PP) cleavage site between the two protein-encoding sequences. Sequencing confirmed what was subsequently deposited at TriTrypDB (Tb427_090021700) and identified two amino acid differences between TbFUT1 in the 927 and 427 strains (L185V and T232A).
The pGEX6P1-GST-PP-TbFUT1 construct was expressed in Escherichia coli and the fusion protein purified as described in Materials and methods. The identities of the two higher molecular weight bands (Figure 2-figure supplement 1, lane 8) were determined by peptide mass fingerprinting. The most abundant band was identified as TbFUT1, while the fainter band was identified as a subunit of the E. coli GroEL chaperonin complex. The apparent molecular weight of GST-PP-TbFUT1 chimeric protein (57 kDa) was consistent with the predicted theoretical molecular weight (58.1 kDa).

Recombinant TbFUT1 is active in vitro
The activity of recombinantly expressed GST-TbFUT1 fusion protein was tested by incubation with GDP-[ 3 H]Fuc, as a donor, and a panel of commercially available mono-to octasaccharides ( Table 1) selected from the literature as possible α-1,2-FUT substrates (Li et al., 2008;Wang et al., 1999;Zhang et al., 2010). The effectiveness of each acceptor was evaluated based on the presence/absence and intensities of the TLC bands corresponding to the radiolabelled reaction products ( Figure 2 and Table 1) (FUTs) shows that TbFUT1 (Tb927.9.3600) lacks a conventional N-terminal type 2 membrane protein transmembrane domain (TM, grey box) but contains conserved motifs I-IV (black boxes). A putative TbFUT1 TM (dashed box) overlaps with motif IV but does not align with the eukaryotic FUT TM and is  (Figure 2, lanes 9 and 20). As expected, no products were observed when acceptor oligosaccharides were omitted from the reaction (Figure 2, lane 4). To confirm the detected activities were specific to the recombinant GST-TbFUT1, and not due to some co-purifying endogenous E. coli contaminant, the assay was also performed using material prepared from E. coli expressing the empty pGEX6P1 vector. No transfer of radiolabelled fucose could be observed under these conditions (Figure 2, lanes 5-7).
A band with the same mobility as free fucose was observed in all assay reactions and was considerably stronger in the presence of the GST-TbFUT1 preparation ( Figure 2, lanes 1-4 and 9-22) than when GDP-[ 3 H]Fuc was incubated with reaction buffer alone ( Figure 2, lane 8) or in the presence of most likely part of a cleavable N-terminal mitochondrial targeting sequence (residues 1-31 with cleavage at R30-R31). Sequences used in the alignment: T. cruzi (TcCLB.506893.90), L. major (LmjF01.0100), H. pylori (AAC99764), Homo sapiens FUT2 (AAC24453), and Mus musculus FUT2 (AAF45146). (B) A selection of known and predicted fucosyltransferase protein sequences was aligned using ClustalΩ (Sievers et al., 2011), and the unrooted phylogram shown was generated by iTOL (itol.embl.de) (Letunic and Bork, 2016). Single homologues of the GT11 TbFUT1 were found in all the kinetoplastids (marked in green) and, collectively, these sequences form a clade distant from other fucosyltransferase sequences, including the other kinetoplastid GT10 fucosyltransferases (marked by a dashed line). No TbFUT1 homologues were found in apicomplexan parasites or in Euglena.
*The relative efficiencies of the acceptors as TbFUT1 substrates (+++, ++, +, and -) are based on visual inspection of the intensities of the products bands in Figure 2.
the material purified from the E. coli cells transformed with the empty vector ( Figure 2, lanes 5-7). These data suggest that TbFUT1 has a significant propensity to transfer Fuc to water. Interestingly, one of the substrates (LNB-O-Me; Galβ1,3GlcNAcβ1-O-methyl) suppressed the amount of free Fuc produced in the reaction (Figure 2, lane 21), suggesting that this glycan may bind more tightly to the TbFUT1 acceptor site than the other oligosaccharides tested and thus prevent the transfer of Fuc from GDP-Fuc to water. In vitro high sugar nucleotide hydrolysis activity has been previously described for at least one member of the GT11 family (Zhang et al., 2010). Inverting α-1,2 and α-1,6-FUTs do not usually require divalent cations for their activity (Beyer and Hill, 1980;Kamińska et al., 2003;Li et al., 2008;Pettit et al., 2010). To study the divalent cation dependence of TbFUT1, the assay was repeated in buffer without divalent cations or containing EDTA. No change in activity was observed in either case, indicating TbFUT1 does not require divalent cations for its activity (Figure 2-figure supplement 2).
Finally, given the propensity for TbFUT1 to transfer [ 3 H]Fuc from GDP-[ 3 H]Fuc to water, producing free [ 3 H]Fuc and presumably GDP, we set up a GDP-Glo assay similar to that used for LmjFUT1 (Guo et al., 2021) to monitor the turnover by recombinant TbFUT1 of non-radioactive GDP-Fuc to GDP (see Materials and methods). In this assay, we could see TbFUT1-dependent turnover of GDP-Fuc to GDP in the absence of acceptor substrate (70 ± 2 pmol) and the stimulation of turnover in the presence of LNB acceptor substrate (265 ± 13 pmol), consistent with the results in the radiometric assay. Under the same conditions, there was no detectible turnover of either GDP-Man or GDP-Glc, showing that TbFUT1 is specific, or at least highly selective, for GPD-Fuc as donor substrate.

Characterization of the TbFUT1 reaction product
The glycan reaction products were structurally characterized to determine the anomeric and stereochemical specificity of TbFUT1. Initially, we performed exoglycosidase and/or acid treatment of the radiolabelled reaction products (recovered by preparative TLC) utilizing Lac, LacNAc, and LNB as substrates. The tritium label ran with the same mobility as authentic Fuc after acid hydrolysis of all three products (   To obtain direct evidence, we performed a large-scale activity assay using LNB-O-Me as an acceptor and purified the reaction product by normal phase HPLC. Fractions containing the putative fucosylated trisaccharide product ( Figure 3-figure supplement 2) were pooled and analysed for their neutral monosaccharide content, which showed the presence of Fuc, Gal, and GlcNAc. The purified reaction products were permethylated and analysed by electrospray ionisation mass spectrometry (ESI-MS) ( Figure 3A), which confirmed that the main product was a trisaccharide of composition dHex 1 Hex 1 HexNAc 1 . The MS/MS spectrum was also consistent with the dHex residue being attached to the Hex, rather than HexNAc, residue ( Figure 3B). Subsequently, partially methylated alditol acetates (PMAAs) were generated from the purified trisaccharide product and analysed by gas chromatography  Table 1), may suggest that TbFUT1 can also form Fucα1,2Fuc linkages.
The purified TbFUT1 reaction product was also exchanged into deuterated water ( 2 H 2 O) and analysed by one-dimensional 1 H-NMR and two-dimensional 1 H-ROESY (Rotating frame Overhouser Effect SpectroscopY). The 1 H-NMR spectrum showed a doublet at about 5.1 ppm, consistent with the signal from the proton on the anomeric carbon (H 1 ) of an α-Fuc residue ( Figure 4A).
A characteristic doublet for the anomeric proton of a β-Gal residue was also observed at 4.5 ppm. In the 1 H-ROESY spectrum, a cross-peak (labelled    a) could be observed indicating a through-space connectivity between the H 1 of α-Fuc and the H 2 of a β-Gal, consistent with a Fucα1,2Gal linkage in the TbFUT1 reaction product ( Figure 4B). The chemical shifts that could be clearly assigned by either one-dimensional 1 H-NMR or twodimensional 1 H-ROESY are listed in Table 3. Taken together, these data unambiguously define the structure of the TbFUT1 reaction product with GDP-Fuc and LNB-O-Me as Fucα1,2Galβ1,3GlcNAcβ1-O-CH 3 which, in turn, defines TbFUT1 as having a GDP-Fuc: β-Gal α-1,2-fucosyltransferase activity with an apparent preference for a Galβ1,3GlcNAcβ1-O-R acceptor motif.

Generation of TbFUT1 conditional null mutants in procyclic and bloodstream form T. brucei
Semi-quantitative RT-PCR showed that TbFUT1 mRNA was present in both bloodstream form and procyclic form T. brucei (data not shown). We therefore sought to explore TbFUT1 function in both lifecycle stages by creating TbFUT1 conditional null mutants. The strategies used to generate the mutants are described in Figure 5. The creation of these mutants was possible because genome assembly indicated TbFUT1 to be present as a single copy per haploid genome, and Southern blot analysis using a TbFUT1 probe was consistent with this prediction (Figure 5figure supplement 1). In procyclic cells ( Figure 5, left panel), the first TbFUT1 allele was replaced by homologous recombination with linear DNA containing the puromycin resistance gene (PAC) flanked by about 500 bp of the TbFUT1 5′-and 3′-UTRs. After selection with puromycin, an ectopic copy of TbFUT1, under the control of a tetracycline-inducible promoter, was introduced in the ribosomal DNA (rDNA) locus using phleomycin selection. Following induction with tetracycline, the second allele was replaced with the BSD gene by homologous recombination, generating the final procyclic form ΔTbFUT1::PAC/TbFUT1 Ti / ΔTbFUT1::BSD conditional null mutant cell line (PCF TbFUT1 cKO). In bloodstream form cells ( Figure 5, middle panel), an ectopic copy of TbFUT1 carrying a C-terminal MYC 3 epitope tag under the control of a tetracycline-inducible promoter was first introduced into the ribosomal DNA (rDNA) locus using phleomycin selection. Following cloning and induction with tetracycline, the first TbFUT1 allele was

TbFUT1 is essential to procyclic and bloodstream form T. brucei
Procyclic and bloodstream form TbFUT1 conditional null mutants were grown under permissive (plus tetracycline) or non-permissive (minus tetracycline) conditions. The PCF TbFUT1 cKO cells grown under non-permissive conditions showed a clear reduction in the rate of cell growth after 6 days, eventually dying after 15 days ( Figure 6A).
The BSF TbFUT1 cKO cells grew like wild-type cells under permissive conditions, whether or not the expressed TbFUT1 had a C-terminal MYC 3 tag, and under non-permissive conditions also showed a clear reduction in the rate of cell growth after 2-4 days, dying after 3-5 days ( Figure 6B,C). These growth phenotypes are very similar to those described for procyclic and bloodstream form TbGMD conditional null mutants that cannot synthesize GDP-Fuc under non-permissive conditions (Figure 6figure supplement 1; . This is consistent with the hypothesis that TbFUT1 may be the only enzyme that utilizes GDP-Fuc, or at least that it is the only FUT transferring fucose to essential acceptors. Further evidence that TbFUT1 is essential for procyclic and bloodstream form growth was obtained from northern blots ( Figure 6D,E). These show that TbFUT1 mRNA levels are undetectable for several days after the removal of tetracycline, but that growth resumes only when some cells escape tetracycline control after about 29 days (procyclic form) and 11 days (bloodstream form). Escape from tetracycline control after several days is typical of conditional null mutants for essential trypanosome genes (Roper et al., 2002). Evidence for the expression of the MYC 3 tagged TbFUT1 protein in the BSF TbFUT1-MYC 3 cKO cell line and of unmodified TbFUT1 in the BSF TbFUT1 cKO cell line under permissive conditions is shown in Figure 6F. From a morphological point of view, both procyclic form TbGMD and TbFUT1 conditional null mutants grown under non-permissive conditions showed an increase in average cell volume, due to increased cell length, concomitant with the start of the cell growth phenotype (Figure 6-figure  supplement 2A). However, we were unable to reproduce the flagellar detachment phenotype previously reported for the PCF TbGMD cKO grown in non-permissive conditions , nor was such a phenotype observed in the PCF TbFUT1 cKO parasites, either by scanning electron microscopy or immunofluorescence (IFA) ( was consistent with what has previously been reported for wild-type cells (LaCount et al., 2002). Additionally, we could observe no defect in cell motility in either PCF TbGMD or PCF TbFUT1 cKO grown in non-permissive conditions ( Figure 6-figure supplement 4).

TbFUT1 localizes to the parasite mitochondrion
The BSF TbFUT1-MYC 3 cKO cell line, grown under permissive conditions, was stained with anti-MYC antibodies and produced a pattern suggestive of mitochondrial localization. This was confirmed by co-localization with MitoTracker ( Figure 7A). However, when TbFUT1 was introduced into wild-type cells fused with an HA 3 epitope tag at the N-terminus, either with or without a C-terminal MYC 3 -tag    (constructs pLEW100:HA 3 -FUT1-MYC 3 and pLEW100:HA 3 -FUT1), the tagged protein co-localized with GRASP, a marker of the Golgi apparatus ( Figure 7B,C). In these cases, we suspect that N-terminal tagging has disrupted mitochondrial targeting by obscuring the N-terminal mitochondrial targeting sequence. Indeed, no mitochondrial targeting motif was predicted in silico for N-terminal HA 3 tagged TbFUT1. Nevertheless, since the mitochondrial localization of a FUT is unprecedented, we elected to raise polyclonal antibodies against recombinant TbFUT1 to further assess its subcellular location. To do so, an N-terminally His 6 -tagged Δ 32 TbFUT1 protein was expressed, re-solubilized from inclusion bodies and used for rabbit immunization. The IgG fraction was isolated on immobilized protein-A observed for anti-MYC and MitoTracker, suggesting that TbFUT1-MYC 3 localizes to the mitochondrion. (B, C) IFA of BSF cKO parasites expressing a tet-inducible N-terminally tagged HA 3 -TbFUT1-MYC 3 (B) or HA 3 -TbFUT1 (C) after labelling with anti-HA, anti-GRASP, and DAPI suggests a Golgi apparatus location for both HA 3 -TbFUT1-MYC 3 and HA 3 -TbFUT1. The absence of anti-MYC (A) or anti-HA (B, C) staining under non-permissive conditions confirms the specificity of the labelling for the respective TbFUT1 fusion proteins. White boxes: areas magnified 1.5× in the rows below. Scale bars: 3 μm.   To further ensure mono-specificity of the antibodies to TbFUT1, the resulting fraction was adsorbed against a concentrated cell lysate of the PCF TbFUT1 cKO mutant grown for 9 days under non-permissive conditions. The resulting highly specific polyclonal antibody was used to detect TbFUT1 expression in wild-type bloodstream form cells as well as in BSF and PCF TbFUT1 cKO cells under permissive and non-permissive conditions ( Figure 8A-C).
Anti-TbFUT1 antibodies co-localized with MitoTracker staining in the wild-type cells and in the conditional null mutants under permissive conditions. No signal for the anti-TbFUT1 antibodies was seen under non-permissive conditions, confirming the specificity of the polyclonal antibody. Taking the possibility of a dual Golgi/mitochondrial localization into account, TbFUT1 localization was also assessed in bloodstream form cells ectopically expressing TbGnTI-HA 3 as an authentic Golgi marker (Damerow et al., 2014). No co-localization between TbGnTI-HA 3 and anti-TbFUT1 was observed, suggesting that TbFUT1 is either exclusively or predominantly expressed in the parasite mitochondrion ( Figure 8D).

Deletion of TbFUT1 disturbs mitochondrial activity
Bloodstream form wild-type and BSF TbFUT1 cKO cells, grown with and without tetracycline for 5 days, were stained with antibodies to mitochondrial ATPase and with MitoTracker. As expected, ATPase and MitoTracker co-localized in wild-type cells and in the mutant under permissive conditions (Figure 9, top panels). However, under non-permissive conditions the few remaining viable cells showed significantly diminished MitoTracker staining, indicative of reduced mitochondrial membrane potential, and a reduction in ATPase staining, suggesting that TbFUT1 is in some way required for mitochondrial function (Figure 9, lower panels, and Figure 9-figure supplement 1).
TbFUT1 can replace the homologous essential mitochondrial fucosyltransferase in L. major The essentiality and mitochondrial location of the orthologous protein in L. major, LmjFUT1, was reported in Guo et al., 2021. Here we asked whether TbFUT1 could functionally substitute for LmjFUT1 using a plasmid shuffle approach (Guo et al., 2017;Murta et al., 2009). We started with an L. major homozygous fut1-null mutant expressing LmjFUT1 from the episomal pXNGPHLEO vector which additionally expresses GFP (Guo et al., 2017;Murta et al., 2009) and transfected it with an episomal construct pIR1NEO-TbFUT1 (±HA) expressing WT or HA-tagged TbFUT1. Following transfection, parasites were grown briefly in the absence of any selective drug, and single cells were sorted for GFP expression and selected for phleomycin resistance, or GFP loss accompanied by G418 (NEO) resistance ( Figure 10A). In these studies, 85% (82/96) of the GFP-positive (GFP+) clonal lines (positive control) grew when placed in microtiter wells, while 81% (156/192) of the GFP-negative (GFP-) sorted cells likewise grew ( Figure 10A). Control PCR studies confirmed the loss of TbFUT1 in the GFP-clones (not shown). The similar survival of cells bearing Lmj or TbFUT1 provides strong evidence that the TbFUT1 can fully satisfy the L. major FUT1 requirement (Murta et al., 2009). Immunofluorescence microscopy of progeny Lmj fut1-/pIR1NEO-TbFUT1 (±HA)-transfected parasites with either mouse monoclonal anti-HA antibodies or affinity-purified rabbit polyclonal anti-TbFUT1 antibodies showed that TbFUT1 was faithfully localized in the Leishmania mitochondrion ( Figure 10B). The Pearson r correlation for co-localization of the anti-HA or anti-TbFUT1 signals with MitoTracker red was >0.98 in both experiments. In a negative control where anti-TbFUT1 antibody was omitted, the fluorescence signal was lost (data not shown). This result establishes conservation of mitochondrial targeting of TbFUT1 when heterologously expressed in L. major promastigotes.

Discussion
The presence and essentiality of the nucleotide sugar donor GDP-Fuc in BSF and PCF trypanosomes led us to search for putative FUTs in the parasite genome. Only one gene (TbFUT1; Tb927.9.3600), belonging to the CAZy GT11 family, was found, and phylogenetic analyses revealed that one, or two in the case of T. cruzi, orthologue could be found in the genomes of other kinetoplastids. These putative kinetoplastid FUTs form a distinct clade within the GT11 FUT superfamily and are distinct from the GT10 FUTs found in T. cruzi, L. major, and related parasites, which are absent in T. brucei. Consistent with TbFUT1 being the only enzyme likely to utilize the essential metabolite GDP-Fuc, we found that TbFUT1 is also essential to both BSF and PCF parasites. We were able to express TbFUT1 in E. coli, and the recombinant enzyme was used to demonstrate its activity as a GDP-Fuc: β-Gal α-1,2fucosyltransferase with an apparent preference for a Galβ1,3GlcNAcβ1-O-R acceptor motif out of the acceptor substrates investigated.
Immunofluorescence analysis, using an affinity-purified antibody raised against native TbFUT1 as well as C-terminal epitope tagging, suggests that TbFUT1 is located in the parasite mitochondrion. Consistent with this, TbFUT1 was identified in the mitochondrial proteome of PCF T. brucei (Panigrahi et al., 2009). Although this was based on a single unique peptide, and is therefore not a high-confidence assignment, this is the only proteomic dataset deposited on TriTrypDB (Aslett et al., 2010) that identified TbFUT1. A quantitative proteomic analysis that estimated protein turnover in T. brucei (Tinti et al., 2019), and ranking of protein abundances using those deep proteomic data sets (Ji et al., 2021), did not identify TbFUT1 in either BSF or PCF, suggesting that it is a very lowabundance protein. The immunofluorescence analysis performed in this study does not allow us to define the specific localization of TbFUT1 in the parasite mitochondrion. Generally, proteins that are targeted to the mitochondrion via an N-terminal presequence are translocated across both the outerand inner-mitochondrial membranes into the mitochondrial matrix where the presequence is proteolytically cleaved (Long et al., 2008). Since TbFUT1 has a predicted mitochondrial targeting sequence ( Figure 1A), we speculate that TbFUT1 may be located in the mitochondrial matrix. However, we are mindful that the putative presequence also overlaps with the predicted FUT motif IV (Li et al., 2008). Future work will address whether the FUT is indeed imported to the mitochondrial matrix, the intermembrane space, or the organelle outer membrane and will include the functional characterization of the role of the predicted TbFUT1 mitochondrial targeting sequence in protein localization. These studies will also assess the fate of motif IV and its function in TbFUT1 enzymology.
Regardless of the precise sub-organellar localization, the localization of TbFUT1 to the mitochondrion is a highly unusual result. Although in recent years fucosylation has been described in the nucleus and cytosol of protists and plants (Bandini et al., 2016;Rahman et al., 2016;Van Der Wel et al., 2002;Zentella et al., 2017), we are unaware of any other examples of mitochondrial FUTs in any organism. Mitochondrial glycosylation in general is poorly understood. The only other known example of a mitochondrial-localized glycosyltransferase is mitochondrial O-GlcNAc transferase (mOGT) (Bond and Hanover, 2015). Mitochondria of rat cardiomyocytes are also positive for OGA and express a UDP-GlcNAc transporter on their outer membrane, indicating all of the molecular components required for this cycling post-translational modification are present in the organelle (Banerjee et al., 2015). Interestingly, disruption of mOGT in HeLa cell mitochondria also leads to mitochondrial dysfunction. While there are no (m)OGT orthologues in kinetoplastids, these observations highlight some of the challenges inherent for a mitochondrial-localized glycosyltransferase. Firstly, for a TbFUT1 that localizes to either the intermembrane space or the mitochondrial matrix to be active, GDP-Fuc would need to be imported into the mitochondrion, which would require the presence of an uncharacterized mitochondrial GDP-sugar transporter. Secondly, TbFUT1 appears to be an α-1,2-FUT that decorates glycans terminating in Galβ1,3GlcNAc, suggesting either that additional uncharacterized glycosyltransferases and nucleotide sugar transporters may be present in the parasite mitochondrion, or that a glycoconjugate substrate may be assembled in the cytoplasm or secretory pathway and then somehow translocated to or imported into the mitochondrion to be modified by TbFUT1. Experiments to resolve these options will be undertaken, as will further experiments to try to find the protein, lipid, and/or other acceptor substrates of TbFUT1. The latter may then provide clues as to why TbFUT1 is essential for mitochondrial function and parasite growth. Several attempts to identify TbFUT1 substrates have failed so far, including fucose-specific lectin blotting and pull-downs, [ 3 H]fucose labelling of parasites transfected with GDP-Fuc salvage pathway enzymes, and LC-MS/MS precursor ion and neutral-loss scanning methods. Although the significance is unclear, it is interesting to note that procyclic TbFUT1 has been shown to be under circadian regulation (Rijo-Ferreira et al., 2017).
In conclusion, TbFUT1 is an essential -1,2-FUT that localizes to either outer membrane or inner compartments of the parasite mitochondrion and presents orthologues throughout the kinetoplastida. As described in Guo et al., 2021, both activity and mitochondrial localization of the L. major homologue are required for parasite viability, whereas other putative L. major FUTs targeted to the secretory pathway are dispensable. Although no data are available so far on the enzymes from T. cruzi or other members of this group, these initial results suggest the intriguing possibility of an essential, conserved mitochondrial fucosylation pathway in kinetoplastids that might be exploitable as a common drug target.  (Wirtz et al., 1999). Procyclic form T. brucei was cultured in SDM-79 medium (Brun and Schönenberger, 1979) containing 15% fetal bovine serum and GlutaMAX at 28 °C. Bloodstream form T. brucei was grown in HMI-9t at 37 °C, 5% CO 2 . Induction was performed at 0.5 μg/ml tetracycline in both forms.

BLASTp searches
The T. brucei, T. cruzi, and L. major predicted proteins (from the GeneDB database, http://www. genedb. org) were searched for putative FUTs using the BLASTp search algorithm (Altschul et al., 1997). The query input sequences are listed in Supplementary file 1. Protein sequence multiple alignments were assembled using Clustal (Sievers et al., 2011) and Jalview (Waterhouse et al., 2009).

Cloning, protein expression, and purification of TbFUT1
The open reading frame (ORF) was amplified by PCR from T. brucei strain 427 genomic DNA and cloned into the N-terminal GST fusion vector pGEX-6P-1, modified to contain the PreScission Protease site (kind gift of Prof. Daan Van Aalten) using primers P1 and P2 (Supplementary file 2). Either the resulting pGEX6P1-GST-PP-TbFUT1 or pGEX6P1-GST-PP were transformed into BL21 (DE3) E. coli strain. Recombinant protein expression was induced with 0.1 mM isopropyl-β-D-thiogalactopyranoside (IPTG) and carried out at 16 °C for 16 hr. Prior to lysis by French press, cells were incubated for 30 min on ice in 50 mM Tris-HCl, 0.15 M NaCl, 1 mM DTT, pH 7.3 (Buffer A) with EDTA-free Complete Protease Inhibitors Tablet (Roche) and 1 mg/ml lysozyme. The soluble fraction, obtained by centrifugation at 17,000 × g, 4 °C for 30 min, was incubated 2 hr at 4 °C with Glutathione Sepharose Fast Flow beads that had been pre-equilibrated in Buffer A. The mixture was loaded into a disposable column and the beads washed first with 50 mM Tris-HCl, 0.25 M NaCl, 1 mM DTT, pH 7.3 (Buffer B), then with 50 mM Tris-HCl, 0.15 M NaCl, 0.1% sodium deoxycholate, 1 mM DTT, pH 7.3 (Buffer C), before being re-equilibrated in Buffer B. GST-TbFUT1 or GST were eluted in 50 mM Tris-HCl, 0.15 M NaCl, 10 mM reduced glutathione pH 8.0 (Buffer E). The yield of recombinant protein was estimated by Bradford assay and the degree of purification by SDS-PAGE analysis. Recombinant protein identification by peptide mass fingerprinting was performed by the Proteomic and Mass Spectrometry facility, School of Life Sciences, University of Dundee.

Fucosyltransferase activity assays Radioactive assay
Aliquots of 2 μg of affinity-purified GST-TbFUT1 were incubated with 1 μCi GDP[ 3 H]Fuc (American Radiochemicals), 1 mM acceptor in 50 mM Tris-HCl, 25 mM KCl, 5 mM MgCl 2 , 5 mM MnCl 2 , pH 7.2 for 2 hr at 37 °C. The acceptors tested (Table 1) were purchased from Sigma, Dextra Laboratories, or Toronto Research Chemicals. To study the dependency on divalent cations, MgCl 2 and MnCl 2 were removed from the buffer and a formulation with 10 mM EDTA was also tested. Reactions were stopped by cooling on ice then desalted on mixed bed as follows: 100 μl each Chelex100 (Na + ) over Dowex AG50 (H + ) over Dowex AG3 (OH -) over QAE-Sepharose A25 (OH -). The flow through and the elutions (4 × 400 μl of water) were combined. About 5% of the desalted reactions were counted at a LS 6500 scintillation counter (Beckmann). The remaining material was lyophilized for further analyses.

Luminescence assay
Recombinant GST-TbFUT1 fusion protein was expressed in E. coli as described above, except that cell lysis was performed without lysozyme using a Cell Disruptor (Constant Systems) in place of a French press. After lysis, the suspension was centrifuged at 45,000× g, 30 min, 4 °C, and the fusion protein was purified using glutathione Sepharose 4B. SDS-PAGE analysis of the glutathione-eluted purified GST-TbFUT1 was like that shown in (Figure 2-figure supplement 1). The purified GST-TbFUT1 was diluted 1:10 into assay buffer and 20 μl aliquots containing about 0.1 μg of GST-TbFUT1 (and no-enzyme controls) were incubated with GDP-Fuc, GDP-Man, or GDP-Glc (final concentrations 50 mM Tris-HCl, pH 7.2, 25 mM KCl, 50 μM GDP-sugar), with and without 0.8 mM LNB, overnight at room temperature in 96-well flat-bottom luminometer plates (Corning Costar 3688). Reactions were stopped and developed with 25 μl Promega GDP-Glo reagent and after 1 hr luminescence was read on a BMG Labtech Pherostar plate reader. A calibration curve using 25 μl aliquots of GDP from 0 to 25 μM in assay buffer was developed in parallel. All measurements were performed in triplicate.

HPTLC analysis
Reaction products and standards were dissolved in 20% 1-propanol and separated on a 10 cm HPTLC Si-60 plates (Merck) using 1-propanol:acetone:water 9:6:4 (v:v:v) as mobile phase. Non-radiolabelled sugars were visualized by orcinol/H 2 SO 4 staining. In the case of radiolabelled products, the HPTLC plates were sprayed with En 3 hance (PerkinElmer) and visualized by fluorography.

Recovery of radiolabelled products
The reaction products from the TbFUT1 activity assays using Lac, LNB, and LacNAc as acceptors were separated on a 20 cm HPTLC Si-60 plate. The radiolabelled products were localized on the HPTLC plate using an AR-2000 (BioScan) plate reader, the corresponding areas were wetted with mobile phase, and the silica scraped and transferred to microcentrifuge tubes. The radiolabelled materials were extracted from the solid phase by incubating with mobile phase and a 5% aliquot counted at the scintillation counter to determine the amount of recovered material.

Acid hydrolysis
Aliquots corresponding to 15% of the extracted radiolabelled reaction products were dried in a Speedvac concentrator, re-dissolved in 4 M TFA, and incubated at 100 °C for 4 hr (Ferguson, 1992). After cooling, the samples were dried on a Speedvac concentrator, washed in water, and re-dissolved in 20% 1-propanol to be analysed by HPTLC as described in the main text.

Xanthomonas manihotis α-1,2-fucosidase reaction
Aliquots corresponding to 15% of the extracted radiolabelled reaction products were treated or mock treated with 10 U X. manihotis α-1,2-fucosidase (New England Biolabs) at 37 °C for 16 hr. The reactions were stopped by heating at 100 °C for 10 min and desalted on mixed bed columns as described for the activity assay. After freeze-drying, samples were washed with water, re-dissolved in 20% 1-propanol, and analysed by HPTLC.
Large-scale TbFUT1 assay and product purification Acceptor (5 mM lacto-N-biose-β-O-methyl) and donor (2.5 mM GDP-Fuc) were incubated with 8 μg affinity-purified GST-TbFUT1 in 20 mM Tris-HCl, 25 mM KCl, pH 7.2 at 37 °C for 24 hr. The reaction products were desalted on a mixed-bed column and lyophilized (see above). The trisaccharide product was then isolated by normal phase liquid chromatography. The reactions were dissolved in 95% acetonitrile (ACN) and loaded on a Supelco SUPELCOSYL LC NH 2 HPLC column (7.5 cm × 3 mm, 3 μm) using an Agilent 1,120 Compact LC system. The purification was performed at 40 °C with a 0.3 ml/min flow rate. The column was equilibrated in 100% Buffer A (95% ACN, 5% H 2 O) for 5 min, before applying a first gradient from 0% to 35% Buffer B (95% H 2 O, 5% ACN, 15 mM ammonium acetate) over 25 min followed by a second gradient from 35% to 80% Buffer B over 20 min. The column was then re-equilibrated in 100% Buffer A for 10 min. About 1.5% of each fraction was potted onto a silica HPTLC plate and stained with orcinol/H 2 SO 4 staining to identify the sugar-containing fractions. An aliquot from each orcinol-positive fraction was then analysed by HPTLC, and the ones containing the putative trisaccharide product were pooled and freeze-dried.

Permethylation, ESI-MS analysis, and GC-MC methylation linkage analysis
Purified TbFUT1 reaction product was dried and permethylated by the sodium hydroxide method as described in Ferguson, 1992. Aliquots were used for ESI-MS and the remainder was subjected to acid hydrolysis followed by NaB 2 H 4 reduction and acetylation (Ferguson, 1992). The resulting PMAAs were analysed using an HP6890 GC System equipped with an HP-5 column linked to a 5975C mass spectrometer (Agilent). For ESI-MS and ESI-MS/MS, the permethylated trisaccharide was dried and re-dissolved in 80% ACN containing 0.5 mM sodium acetate and loaded into gold-plated nanotips (Waters) for direct infusion. The Q-Star XL mass spectrometer equipped with Analyst software (Applied Biosystems) was operated in positive ion mode using an ion spray voltage of 900 V and a collision voltage of 50-60 V in MS/MS mode.

NMR
The purified TbFUT1 reaction product was exchanged in 2 H 2 O by freeze-drying and analysed by onedimensional 1 H-NMR and two-dimensional 1 H-ROESY. All spectra were acquired on a Bruker Avance spectrometer operating at 500 MHz with a probe temperature of 293 K.

Generation and expression of epitope-tagged TbFUT1 constructs
TbFUT1 was introduced in two different sites of pLEW100HXM to yield HA 3 -TbFUT1 and HA 3 -TbFUT1-MYC 3 . Starting from the pLEW100 vector, we inserted a 99 bp adapter oligo encoding for the following sequence HindIII/NdeI/AscI/BbvCI/XbaI/BstBI/XhoI/HA/PacI/BamHI, where HA stands for a single HA tag (O1 and O2 in Supplementary file 2). The HA tag was then replaced with a MYC 3 tag amplified from a pMOTag43M plasmid, creating the following multiple cloning site: HindIII/NdeI/As-cI/BbvCI/XbaI/BstBI/XhoI/MYC 3 /PacI/BamHI. This vector was given the name pLEW100XM. TbFUT1 was amplified from a plasmid source using primers P15/P16 and inserted via HindIII/XhoI. pLEW100XM was then used to introduce an HA 3 adapter which would allow N-terminal tagging of a protein of choice. The adapter consisted of the O3/O4 primer pair (Supplementary file 2). The primer pair was fused and inserted into HindIII/AscI restriction sites of pLEW100XM creating the following multiple cloning site: HindIII/HA 3 /AscI/BbvCI/XbaI/BstBI/XhoI/MYC 3 /PacI/BamHI. This plasmid was termed pLEW100HXM. TbFUT1 was amplified using primers P21/P22 for HA 3 -TbFUT1 and P23/P16 for HA 3 -TbFUT1-MYC 3 (Supplementary file 2). The two plasmids were purified and electroporated into BSF cells as described below.

Generation of TbFUT1 conditional null mutants
About 500 bp of the 5′ and 3′ untranslated regions (UTRs) immediately flanking the TbFUT1 ORF were amplified from T. brucei 427 genomic DNA (gDNA) using primers P3/P5 and P4/P6 (Supplementary file 2) and linked together by PCR, and the final product was ligated into pGEM-5Zf(+). Antibiotic resistance cassettes were cloned into the HindIII/BamHI restriction sites between the two UTRs to generate constructs either containing puromycin acetyltransferase (PAC) or blastacidin S deamidase (BSD). To allow cloning of the antibiotic resistance cassettes using HindIII/BamHI, a BamHI site present in the 5′-UTR was first mutated to ggctcc using the QuickChange Site-Directed Mutagenesis Kit (Stratagene) and primers P7/P8. These two constructs were used to replace the two gene copies in procyclic form parasites. In addition, a hygromycin (HYG)-based TbFUT1 gene replacement cassette was generated with longer UTRs (1.25 kb) amplified using primers P9/P10 (5′ UTRs) and P11/P12 (3′UTRs). To avoid the endogenous 5′ UTR site, first the 3′UTR was introduced via BamHI/SacI then the 5′UTR via SpeI/HindIII. The HYG resistance cassette replaced both copies of TbFUT1 in bloodstream form parasites. The tetracycline-inducible ectopic copy construct was generated by amplifying the TbFUT1 ORF from T. brucei gDNA (primers P13/P14) and cloning the resulting PCR product into pLEW100. Additionally, pLEW100XM was generated which allowed universal tagging of a protein of interest with a C-terminal MYC 3 tag (see details below). Linearized DNA was used to transform the parasites as previously described (Burkard et al., 2007;Güther et al., 2006;Wirtz et al., 1999). The genotype of the transformed parasites was verified by Southern blot as detailed below. The TbFUT1-MYC 3 cell line was used as parental strain for the BSF TbFUT1 cKO cell line by eliminating the MYC 3 tag and inserting a cassette comprised of the native (untagged) TbFUT1 gene, a short intergenic region (IGR), and the PAC resistance gene via homologous recombination. A C-terminal fragment of the TbFUT1 ORF (nucleotides 444 to end) was amplified with primers P28/29 (Supplementary file 2) and cloned into the KpnI/ApaI restriction sites of pMOTag4YH-PAC (Oberholzer et al., 2006), a kind gift of Prof. Thomas Seebeck (University of Bern). The eYFP-3xHA cassette encoded in the plasmid was removed by XhoI/SalI digestion followed by re-ligation. Finally, the rDNA homology sequence was digested with NotI/SacI from pLEW100HXM and cloned into these same sites downstream of the puromycin resistance gene (PAC), resulting in pMOTag-TbFUT1-IGR-PAC-rDNA. The linearized plasmid was electroporated in BSF TbFUT1-MYC 3 cKO. Homologous recombination via the TbFUT1 C-terminus and rDNA sequences resulted in generation of BSF TbFUT1 cKO ( Figure 5C).

Southern blotting
Genomic DNA was prepared from 5 × 10 7 or 1 × 10 8 cells using DNAzol (Helena Biosciences) and digested with 0.1 mg/ml RNAse I. The gDNA (5 μg/lane) was digested with the appropriate restriction endonucleases and probed with the ORFs of TbFUT1, puromycin acetyltransferase gene (PAC), hygromycin B phosphotransferase (HYG), or blastacidin S deamidase gene (BSD) generated using the PCR DIG Probe Synthesis Kit (Roche). The blot was developed according to the manufacturer's instructions.

Northern blotting
Total RNA was prepared from 5 × 10 6 to 1 × 10 7 cells using the RNeasy MIDI Kit (Qiagen) according to the manufacturer's instructions. The RNA was separated on a 2% agarose-formaldehyde gel, blotted, and detected using the Northern Starter Kit (Roche). Probes were designed based on the DIG RNA Labelling Kit T7 (Roche), and TbFUT1 and alpha-tubulin (Tb427.01.2340) templates were amplified from T. brucei bloodstream form gDNA using primers P17/P18 and P19/P20 (Supplementary file 2), respectively. Total RNA and DIG labelled probes were quality checked by capillary electrophoresis on an Agilent BioAnalyzer 2100.

Growth curves
Growth of parasites lacking TbFUT1 was analysed by determining cell counts for three clones each of PCF TbFUT1 cKO and BSF TbFUT1 cKO with either tagged of untagged TbFUT1. All cell lines were washed in media without tetracycline (tet) and inoculated at a concentration of 5 × 10 5 parasites/ml (procyclic form) and 2 × 10 4 parasites/ml (bloodstream form). Parasites were grown plus and minus tetracycline and in biological duplicates (PCF) or triplicates (BSF). Cells were counted every 24 hr using a CASY Cell Counter + Analyser system. To keep parasites in log phase, cultures were diluted back to the initial inoculum concentration every 48 hr.

Preparation of anti-FUT1 antibody
The NdeI site within the TbFUT1 sequence was eliminated using the QuickChange Lightning Site-Directed Mutagenesis Kit (Agilent) and primers P24/P25. The resulting mutagenized ORF was then amplified with primers P26/P27 and cloned into pET15b, resulting in an N-terminally truncated construct encoding Δ 32 TbFUT1 fused to an N-terminal hexahistidine tag (HIS 6 ) with PreScission plus (PP) protease cleavage site. E. coli BL21 gold (DE3) cells were allowed to express His 6 -PP-TbFUT1 overnight at 25 °C in auto-inducing media (5052-NPS-MgSO 4 ). Cells were harvested by centrifugation and incubated for on ice in 50 mM Tris-HCl pH 7.3, 0.15 M NaCl, 1 mM DTT (Buffer A) with EDTAfree Complete Protease Inhibitors Tablet (Roche), before being disrupted in a French Press. Since all of the protein was insoluble within inclusion bodies, a refolding procedure was applied. The cell lysate was spun down for 20 min at 20,000 × g and the pellet was washed in ice-cold PBS with EDTAfree Complete Protease Inhibitors Tablet (Roche). Cell debris layering the inclusion body pellet was scratched off and the pellet washed again. This process was repeated three times. The inclusion body pellet was washed with PBS + 0.5% TX-100 and sonicated on ice. After digestion of residual DNA and RNA, the pellet was washed with PBS + 0.1% Tween-20, sonicated, and washed in TBS. The washed inclusion bodies were resuspended in 8 M urea/TBS and incubated overnight at 37 °C, 200 rpm. Insoluble material was spun down and the supernatant slowly diluted down under agitation to 2 M urea with TBS at room temperature. After centrifugation, the supernatant concentrated to 2 mg/ml in a 30 kDa cut-off VivaSpin concentrator. Re-solubilized protein (2 mg) was sent to DC Biosciences Ltd for production of polyclonal rabbit antiserum. The IgG fraction from the hyperimmunized rabbit serum was purified on 2 × 1 ml HiTrap Protein G HP columns (GE Healthcare) and then affinity-purified on GST-TbFUT1 fusion protein coupled to a HiTrap NHS-Activated HP column (GE Healthcare). Finally, the IgG fraction was adsorbed against PCF TbFUT1 cKO cell lysates prepared on day 9 without tetracycline, a time point at which they are virtually free of TbFUT1. The cell lysate was prepared by sonication of 2 × 10 8 cells on ice in 100 mM sodium phosphate buffer pH 7 including EDTA-free Complete Protease Inhibitors (Roche) and 0.01% TX-100. After addition of purified TbFUT1-specific IgGs to the cell lysate, the solution was incubated rotating overnight at 4 °C, concentrated in a VivaSpin 30 kD cut-off filter, and stored in 50 % glycerol at -20 °C. Final estimated rabbit anti-TbFUT1 IgG concentration was 50 µg/ml.

Immunofluorescence microscopy
Late log phase T. brucei bloodstream form or procyclic cells were fixed in 4% PFA/PBS in solution at a concentration of 5 × 10 6 cells/ml. When using MitoTracker Red CMX Ros, cell cultures were spiked with a 25 nM concentration over 20 min before harvesting. Fixed parasites were permeabilized in 0.5% TX-100 in PBS for 20 min at room temperature before blocking in 5 % fish skin gelatine (FSG) containing 0.1% TX-100 and 10% normal goat serum. Alternatively, procyclic cells were permeabilized in 0.1% TX-100 in PBS for 10 min and blocked in 5% FSG containing 0.05% TX-100 and 10% normal goat serum. Primary antibody incubations were performed in 1% FSG, 0.1% TX-100 in PBS. The following primary antibodies were used in 1:1000 dilutions: rabbit anti-TbFUT1 IgG (see above), rabbit anti-GRASP IgG (gifts from Prof. Graham Warren, University College London, and Prof. Chris de Graffenreid, Brown University), rabbit anti-ATPase IgG (gift from Prof. David Horn, University of Dundee), mouse anti-HA IgM (Sigma), rabbit anti-HA IgG (Invitrogen), and mouse anti-MYC IgM (Millipore). Anti-FAZ was used at 1:2 and anti-PFR at 1:10 (both antibodies were kind gifts of Prof. Keith Gull, University of Oxford). After washing, the coverslips were incubated with a 1:500 dilution of respective secondary Alexa Fluor goat IgG, like anti-mouse 488 and 594, as well as anti-rabbit 488 and 594 (Life Technologies). Coverslips were mounted on glass slides with Prolong Gold DAPI antifade reagent (Invitrogen). Microscopy was performed on a DeltaVision Spectris microscope (GE Healthcare), and images were processed using Softworx. For Figure 6-figure supplement 3, microscopy was performed on a Zeiss LSM 700 confocal microscope. All images were further processed using Fiji (Schindelin et al., 2012).

Scanning electron microscopy
Samples were prepared and analysed as described in  Briefly, PCF TbFUT1 cKO cells were grown in the presence or absence of tetracycline and fixed on days 6 and 12 in SDM-79 with 2.5% glutaraldehyde, and the samples were further prepared by the Dundee Imaging Facility, School of Life Sciences. Cells were collected on 1 μM Shandon Nuclepore membrane filters, rinsed twice in 0.1 M PIPES buffer pH 7.2, and treated with 0.2% osmium tetroxide overnight. Samples were then washed in water and in a gradient of ethanol solutions to dehydrate them. After critical point drying on a BALTec critical point dryer D30, the samples were mounted on aluminium stubs that were then coated with 40 nM gold/palladium. The stubs were examined in a Philips XL30 environmental scanning electron microscope operating at an accelerating voltage of 15 kV.

Cell volume determination
The average cell volume was obtained by analysing PCF TbFUT1 cKO and TbGMD cKO on a CASY Cell Counter + Analyser system. Procyclic form cultures grown in the presence or absence of tetracycline were diluted 1:10 in 1× PBS before being diluted in the CASY-ton solution for measurements (>4000 cells/measurement). Two biological replicates were analysed for procyclic form grown in the absence of tetracycline. The p values were calculated performing a two-sided t-test in RStudio.

Sedimentation assay
The assay was adapted from Bastin et al., 1999. Briefly, about 5 × 10 6 T. brucei PCF TbFUT1 cKO or TbGMD cKO cells were resuspended in 1 ml of SDM-79 medium and transferred into a 2 ml polystyrene cuvette (Sarstedt). The cuvettes were incubated at 28 °C (CPS-controller, Shimadzu) in a UV-1601 Spectrophotometer (Shimadzu) and the optical density at 600 nm measured every 30 min. SDM-79 medium was used as a blank. Two biological replicates were analysed for each cell line grown in the presence or absence of tetracycline.

L. major complementation by 'plasmid shuffling'
For the plasmid swap experiments described in Figure 10, the TbFUT1 ORF was amplified by PCR and inserted into the expression vector pIR1NEO yielding pIR1NEO-TbFUT1 (B7148). L. major parasites were grown as promastigotes in M199 medium, transfected, and plated on selective media for clonal lines as described previously (Guo et al., 2017). 'Plasmid shuffling' to swap an episomal LmjFUT1expressing construct with the ones expressing-TbFUT1 (±HA tag) was performed as described in the text and Figure 10A, following procedures described in Guo et al., 2017.

Additional files
Supplementary files • Supplementary file 1. Summary of BLASTp searches for putative fucosyltransferases in the three kinetoplastids genomes.
• Supplementary file 2. Primers and oligonucleotides used in this study.
• Source data 1. This folder includes the raw data for the following figures: Figure  • Source data 2. This folder includes the raw data for Figure 5-figure supplement 1.
• Source data 3. This folder includes the raw data for Figure 6D-F.

• Transparent reporting form
Data availability All data is available in the manuscript. No sequencing, proteomics or protein structural data were generated.